The original title is: "AgentScaffold: Memory, Peer Review, and Continuous Improvement for AI Coding Agents"

Short summary

AgentScaffold is a Python MCP server that adds governance to AI coding agents via durable memory, peer review, and continuous improvement. It parses repositories into a DuckDB property graph with Tree-sitter, indexing code alongside governance artifacts like plans, ADRs, and review findings. The author, a former industrial engineer turned data scientist, applies systems-level process design to the problem of agents that confidently produce uncalibrated output without learning from past mistakes.

• •AgentScaffold provides memory, peer review, and continuous improvement for AI coding agents via MCP
• •Parses repos into a DuckDB+DuckPGQ property graph supporting 8 languages with ~20 node types and ~40 edge types
• •Works with Cursor, Claude Code, Windsurf; ships as pip-installable Python package with CLI

Generated with AI, which can make mistakes.